Anthropic Investigates Unauthorized Access to Mythos AI

Anthropic, the US tech startup behind the Claude chatbot, is investigating claims that unauthorized users accessed its advanced AI model, Mythos. According to The Guardian, the company confirmed on Wednesday it was probing reports that a group of people gained unauthorised access to Mythos through a third-party vendor environment.

Mythos represents a serious potential threat to global cybersecurity as it can identify and exploit 'zero-day' vulnerabilities in major operating systems and web browsers. These flaws are unknown to developers, making them particularly dangerous if exploited by hackers. The UK's National Cyber Security Centre (NCSC) has urged companies to replace obsolete tech due to the urgency posed by such AI threats.

South Korea's intelligence agency has also warned of emerging AI-driven cyberattacks as a potential 'game changer.' According to UPI, the National Intelligence Service issued a government-wide security advisory after identifying growing risks tied to advanced AI models like Mythos. The agency noted that these tools can autonomously detect vulnerabilities and execute hacking operations, marking a shift from AI assisting hackers to AI acting as the hacker itself.

The unauthorized access to Mythos raises concerns about the pace of development and the ability of tech companies to keep their riskiest products out of the public domain. While Anthropic has allowed select tech firms and banks to assess the model's risks, there is ongoing speculation about its potential impact on global cybersecurity. The UK government and regulators are closely monitoring the situation, with Mythos added to high-level discussions at the Cross Market Operational Resilience Group meetings.